No! It is a common misconception that ensemble Kalman approaches require Gaussian prior distributions for them to work.
The ensemble Kalman based method is the optimal method (in a mean squared error sense) under Gaussian assumption, but this does not mean that it does not work for non-Gaussian distributions. The only formal requirement is that the distribution is continuous. However, it is important to note that the performance of the ensemble Kalman methods are not optimal for multi-modal distributions. Furthermore, we typically advise the user to apply various transforms to parameters as part of the data conditioning process in ResX. Examples include log transformation for permeability, or probability transform for probabilities when configuring the model uncertainties in a ResX study.
